waleko/roberta-arxiv-tags

results

This model is a fine-tuned version of FacebookAI/roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0289
• Precision: 0.8165
• Recall: 0.5979
• F1: 0.6903
• Exact Match: 0.4261
• Avg Errors: 0.0085

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Exact Match	Avg Errors
0.0356	1.0	4613	0.0341	0.7847	0.5538	0.6493	0.4041	0.0095
0.0314	2.0	9226	0.0302	0.8274	0.5726	0.6768	0.4310	0.0087
0.0288	3.0	13839	0.0289	0.8165	0.5979	0.6903	0.4261	0.0085

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1